Introduction There is a large volume of scientific works concerning the utility of the 12-item Short-Form Health Survey (SF-12) and its mental health component (MCS-12) for diagnosing depression and anxiety disorders in samples of a general population. Because of the existence of a great number of studies using SF-12 data, it is convenient to estimate depressive symptoms from measurements of MCS-12 in more susceptible populations such as the elderly. In this paper we studied the validity of this scale as a screening measure of severe depressive symptoms in epidemiological research.

Objectives To elect a cut-off point in the MCS-12 scale leading to proper discrimination of severe cases of depressive symptoms as measured by the GDS scale, with good sensibility and specificity.

Methods Data come from part of the SABE study—a longitudinal research in the city of São Paulo, 2000/2006/2010. N=1155 people aged 65 and plus were interviewed in the 2006 round. ROC analysis was used to select a cut-off point in the MCS-12.

Results 905 persons reported answers to both scales. General prevalence of severe depression was 2.6% (1.6 males, 3.2 females). A cut-off point of 43 in MCS-12 leaded to 88% area under ROC curve. Accuracy was 89.3 (86.5 females, 93.5 males), sensibility 0.73 (0.72 females, 0.75 males) and specificity 0.90 (0.87 females, 0.94 males).

Conclusions The validation parameters found in this research indicate that the MCS-12 is a valid measure of depressive symptoms in epidemiologic studies of elders, and a screening tool for depression in clinical practice.